McDonald’s assistant adds MCP skills

A GitHub contest entry called McButler turns McDonald’s menu data into a set of ordering tools. It offers daily recommendations, checks the nutrition of a meal by breaking it into items, compares store options using price, distance and estimated time, and plans group meals around budgets and preferences. The repository had 120 stars in the harvest. The project page says it connects to McDonald’s MCP server and can also run offline using sample data.

For a business owner, the interesting part isn’t the menu. It’s how one data connection can support several practical tasks, from choosing a nearby option to preparing a group order. McButler also includes a meal health check and a playful feature that stages a short debate among characters using the user’s coupons, score and calorie data. Those are features of a contest project, not evidence of a McDonald’s product or a live service. The project explicitly says it was independently developed and is not an official McDonald’s product.

That’s a useful pattern to watch: connect a tool to company data, then give it specific jobs people already do by hand. The same principle applies to internal operations, where the value depends on the quality of the data and whether the tool handles real-world limits honestly. McButler, for example, says queue time must be entered or estimated. This is the kind of thing we build into an AI command centre, where teams can bring related tools and information together.